gabesmed

Project creator here, this was an experiment inspired by the WitnessIRL thread on reddit to find puzzle patterns from the Witness anywhere. I'm also an experienced web engineer but pretty new to deep learning, so it was an opportunity to learn the concepts while applying to a meaningful problem.

Exactly! The root of all this (I think) is the contracting structure for building jet fighters -- cost plus percentage -- was applied to software, which has such a different cost structure. With that structure there is no motivation to ship working software, or work efficiently at all. So people who value that won't work for any company that behaves like that, hence the shift of talent elsewhere.

The youth helped us mostly because we were able to hop up and live in that house for months in order to be close to CMS - know our customers! But even for us those days are over - we have offices in SF and DC and have multiple more established folks on our team. Just like major tech companies, we want to provide a comfortable environment for skilled engineers of any age and family situation.

node/MySQL/nginx for both app2 and SLS. both are on AWS -- the first federal govt systems to use them. Even with the urgency, needed to wait almost a year for them to be approved, mostly around security concerns.

Changing the government contracting process is going to be the slowest part of this -- yep nothing drastic is different now that a few years ago. But the winds of change are in the air. The fine folks at USDS are working hard to fix this from the root. There was a recent California child welfare software project that was broken up into smaller pieces and served as a prototype of a contract that was made to appeal to modern firms, and reward sound development practices. Expect more of this to shift the playing field and lay the groundwork for competition to do the rest.

We basically rebuilt things from scratch on a node/MySQL/nginx stack while providing the same API as the old system, and were able to swap it out. We have become masters of the strangler pattern of software design...

The core of the issue was that EIDM, the old system, was an assembly of products designed for corporate intranets on the orders of 1000s of users -- it was never meant to be used for a consumer system. So all this super complicated permissions logic was bogging everything down.

They mitigated the effects for a while by running it on bigger computers - when we came in it was running on Oracle's BIGGEST computer, the Exadata. $6MM per environment. Terabytes of ram. But still just completely bogged down by a set of software not designed for its use case.

We love 18f!!! We work a little more closely with USDS as our relationship with them is more complementary - they find opportunities in government and will often consult with us and other modern firms on how to approach big problems like identity management, identity verification, etc. we don't work directly with 18F but feel very much like we are part of the same movement towards reforming the ecosystem

We had the dubious fortune of coming in in a crisis situation, which let us insist on our way of doing things. (Bringing own computers, AWS, modern stack, quick hiring, etc)

Even then, though, launching our first product (App 2.0) to production was really uphill as we fought to earn the trust of CMS and the other contractors, who didn't really believe we could deliver.

But after we shipped, and people saw that we could deliver working software, things got way easier. Now we are consulted on major software architecture decisions and are a key part of the design process.

It goes the other way round though as well -- we have learned you have to respect your partners and champions on the inside, and conform a bit as well. When you are dealing with really sensitive personal info you can't play as fast and loose as you could in a brand new startup.

we are at a point now where we can often look at the underlying need and suggest alternate ways of approach. For instance after a big breach a few months ago security wanted us to use some VPN for all our prod systems- we managed to compromise on amazons MFA which is quite solid but gets us the enhanced security they were going for

Long term we want to help reform contracting so that requirements are not framed in this lots-of-boxes-to-check-none-having-to-do-with-serving-users mindset

A lot of the most effective pushback we end up never seeing at all -- it happens in decisions and conversations we aren't party to. We do think in the long run delivering a radically better product will win out... But the presence of pushback does really demonstrate the need for tech-savvy partners inside the government, like those we have been lucky to work with at the Center for Medicare and Medicaid Services, and a reformed contracting process. The only way there will be better technology in government long term is if the people signing the checks are able to discern between the modern way and the old way, and insist on a better way of doing things. (Apologies for brief responses - on mobile)

Todd Park, the ex-CTO of the United States, started to process of pulling in Silicon Valley talent. We were called "Todd's kids" for the longest time :)

Todd reached out to the CEO of Civis Analytis (Dan Wagner I believe?), who had pitched in with the Obama campaign. Through him Todd reached Mikey Dickerson, who was at Google at the time (Mikey is now the head of USDS). Mikey looped in a few other googlers...and so it spread.

I found the MPL team through friends in the startup world, many of whom were ex YC founders

Hey all, i'm one of the team members of Nava, the public benefit corporation that emerged from the MPL team to continue to improve how the government serves its people...happy to answer any questions!

Nava - Washington DC

I'm part of a small team of expats from Silicon Valley that came out to DC last year to help fix healthcare.gov. (Mostly Stanford, Google, Y Combinator folks) We've learned a ton, and see an opportunity to create software that radically improves how our government serves its people - we've started our company, Nava, to chase that vision. In the last 6 months, we've launched 2 major projects:

- App 2.0, the new insurance application for healthcare.gov. It's processing 70% of apps coming through the marketplace, and: converts 35% more people than the old app, gets them through in half the time, is mobile-friendly (20% of apps), sits on robust, scalable infrastructure, etc etc. It was the first system that CMS (Centers for Medicare and Medicaid Service) ever hosted on Amazon AWS.

- Scalable Login System (SLS), the new identity management system for healthcare.gov. Sadly, the main achievement here is that the system is actually up and available, unlike its predecessor. It's now saving the taxpayer $70M in annual maintenance costs.

Here's a Wired article from last summer about us: http://www.wired.com/2014/06/healthcare-gov-revamp/

We're just starting on a few major new projects that have a huge potential to improve our government. We have the technical experience to do things the right way, and the relationships and trust inside government agencies to get things done. Our momentum is strong, the problem is important, and we're looking for good people.

If you're the type that would get fired up about reimagining broken government experiences and technology, please drop us a line at jobs@navahq.com, we'd love to hear from you!

Specifically, we’re looking for:

- a product manager with a technical background

- a senior devops engineer

- a senior full-stack engineer

Yeah, this would be great to have, especially for live previewing of different material combinations. Would require a major rev of the interface/rendering though, so didn't make the cut for launch.

It's called Emmet's "The Good Stuff" - a gel urethane that gives a nice warm glow to the wood, is easy to apply, quick-drying, and not unpleasant or dangerous to breathe. (we work out of our cottage, so no professional ventilation systems for us!) http://www.mapleblock.com/detail/butcher-block-finishes-39/

We'd love to make a custom floor! We can do any size piece as long as it's tiled by panels. It'd be cheaper to use the standard sizes (so the largest panel to tile would be 16x20) as it helps us conserve material. But we can conceivably make single maps or panels as large as 18x32, which is the size of our laser-cutter.

We're actively working on an OSM render path. Coming soon!!

Yup. The working title for the project was 'mapquetry'. Though technically inaccurate, Woodcut Maps is way catchier.

Then I would personally salute you. Still trying to convince the fiancee to do our living room as a woodcut floor :)

Our oboist team member, who hand-carves her own reeds, assembles and finishes each map by hand.

I'll be preparing a preso for WhereCamp this weekend; I'll definitely put it online as well.

Heya, I'm the developer behind Woodcut Maps!

The site is built on python/django and hosted on AWS.

Most of the interesting work happens in python with PIL, numpy and scipy. We download maps tiles (Google for now, working on an OSM render path), and split them into channels for landscape, water, roads and parks. Then we apply several steps of scipy's image morphology filters to get the masks ready for laser cutting and to ease assembly by simplifying the design. Removing small pieces, filling small gaps (which gives the maps the distinctive look of dense roads being filled in, creating chunky blocks), and simplifying boundaries.

All the rendering happens on picloud.com, which has been a really robust part of our infrastructure. It helps us scale up and down based on demand without having to manage an autoscaling AWS backend processor setup ourselves, or overpay for AWS instances.

I'm probably going to talk more in detail about it at WhereCamp this weekend if anyone wants to talk in person.

Another example to look at:

http://www.appsafari.com/games/10766/kingpin-game/

Kingpin, ... is a new addition to the foursquare client family. ... Users can accumulate points, collect badges to store in vaults, and become mayor (boss in Kingpin) of places they regularly frequent. .... Kingpin has all the nifty features of foursquare, but it raises the stakes by allowing users to attack local bosses.

I wish I'd known him.